Table 1.
Abbreviations of multiple comparison procedures (and their corresponding controlled error rate), used in the text and in summary figures.
Fig 1.
Average FWER for different methods purporting to control the FWER at α = 0.05.
A can be thought of as the magnitude of differential abundance for truly differentially abundant features, and ρ is the true correlation within blocks of dependent tests. The blue solid circles represent the case of 18 dependent tests (out of 100 total), whereas the red solid triangles are for the case of 36 dependent tests. Parentheses indicate ± 2 SEM.
Fig 2.
Average power for different methods purporting to control the FWER at α = 0.05.
A can be thought of as the magnitude of differential abundance for truly differentially abundant features, and ρ is the true correlation within blocks of dependent tests. The blue solid circles represent the case of 18 dependent tests (out of 100 total), whereas the red solid triangles are for the case of 36 dependent tests. Parentheses indicate ± 2 SEM.
Fig 3.
Average FDR for different methods purporting to control the FDR at α = 0.05.
A can be thought of as the magnitude of differential abundance for truly differentially abundant features, and ρ is the true correlation within blocks of dependent tests. The blue solid circles represent the case of 120 dependent tests (out of 2000 total), whereas the red solid triangles are for the case of 360 dependent tests. Parentheses indicate ± 2 SEM.
Fig 4.
Average power for different methods purporting to control the FDR at α = 0.05.
A can be thought of as the magnitude of differential abundance for truly differentially abundant features, and ρ is the true correlation within blocks of dependent tests. The blue solid circles represent the case of 120 dependent tests (out of 2000 total), whereas the red solid triangles are for the case of 360 dependent tests. Parentheses indicate ± 2 SEM.